广义回归神经网络在显示器颜色空间转换中的应用

目的 研究广义回归神经网络对显示器色彩空间转换预测准确性的方法。方法 通过编程， 借助MeasureTool软件自动测量， 获取建模和测试数据， 通过反复测试选择建模合适的参数， 并用广义回归神经网络模型进行仿真实验， 以获取较好的RGB-Lab色空间转换模型。结果 经广义回归神经网络模型测试， 得到色块平均色差达到2.5275， 最大色差达到19.3620。结论 该方法建模简单方便， 网络训练速度快， 转换精度高， 对显示器颜色空间转换具有较好的非线性拟合能力和更高的预测准确性。

Application of General Regression Neural Network in the Display of Color Space Conversion

This study aimed to investigate the method for accuracy prediction of color space conversion by general regression neural network. Through the modeling, testing data was collected by MeasureTool software of automatic measurement. The test was repeated for optimization of modeling right parameters. Finally, the simulation experiment was carried out using General regression neural network model, and a better RGB to Lab color space conversion model was obtained. Our results showed that the average color difference on testing General regression neural network model reached 2.5275, and the maximum color difference was 19.3620. In general, the established method is easy and convenient, which has good non-linear fitting capability and higher prediction accuracy in color space conversion.